MCP Tools

Describe a file job in plain English and get the Convert Fleet tools that do it, ranked, each with example arguments you can send straight back.

When to use — Any time you are not certain which tool a request maps to. It costs no credit and one call beats guessing wrong twice.

The Convert Fleet manual as Markdown: overview, per-tool reference, multi-step recipes, error handling and connection setup.

When to use — Starting a multi-step file job, or when a tool behaved in a way you did not expect and you want the contract spelled out.

Convert a file between any of Convert Fleet's 178+ supported formats (documents, images, audio, video, e-books, archives, fonts, spreadsheets). Provide the input as a public file_url (preferred) or inline file_base64 + filename, and the target_format (e.g. 'pdf', 'mp3', 'docx', 'webp'). Returns a short-TTL signed download URL for the converted file. Use list_supported_formats first if unsure a target is valid.

When to use — The user has a file in one format and needs it in another: docx to pdf, heic to jpg, wav to mp3, mkv to mp4, epub to mobi, csv to xlsx, svg to png. This is the default tool for any 'turn X into Y' request and covers 178+ formats.

Reduce a file's size while keeping its format and reasonable quality. Best for images (png/jpg/jpeg/webp/gif) and media (mp4/mp3/wav) — the file is re-encoded at a smaller size. Provide file_url or file_base64 + filename.

When to use — A file is too big for an upload limit or an email, and the format must stay the same. Best on png/jpg/jpeg/webp/gif and mp4/mp3/wav.

Run a PDF operation: merge several PDFs, split into pages, rotate, delete-pages, extract-pages, number-pages, watermark, or protect. For 'merge' pass file_urls[] or file_base64s[]; for everything else pass a single file_url or file_base64. 'split' returns the pages as base64; the other operations return a signed download URL.

When to use — Editing the structure of a PDF: merging several into one, splitting into pages, rotating, deleting or extracting a page range, adding page numbers or a watermark, or password-protecting it.

Run FFmpeg on a media file (transcode, trim, scale, extract audio, change frame rate, make a thumbnail, etc.). Pass the input as file_url or file_base64 + filename, the FFmpeg arguments that go AFTER the input in 'args' (e.g. '-vf scale=1280:720 -c:v libx264' or '-vn -ab 192k'), and the output_format (e.g. 'mp4', 'mp3', 'gif'). The server runs: ffmpeg -i <input> <args> output.<output_format>. Shell metacharacters and network/file protocols are blocked.

When to use — A media job the dedicated tools do not cover: transcoding with specific codecs or bitrates, scaling, changing frame rate, extracting the audio track, making a thumbnail, or applying an arbitrary filter graph.

Blur, pixelate or black out part of a video — or all of it. Pass the input as file_url or file_base64 + filename. Set mode='regions' (default) with 'regions' to cover only certain areas, mode='whole' to blur every pixel of every frame, or mode='whole-except' to blur everything BUT the listed areas. 'regions' is a JSON array string of boxes measured as fractions of the frame: [{"x":0.35,"y":0.12,"w":0.2,"h":0.25}] puts a box 35% across and 12% down, a fifth of the frame wide. Per area you can also set shape (rect|ellipse), style (blur|pixelate|fill), strength (1-100), feather (0-100), start/end in seconds, and track — [{"t":0,"x":0.1,"y":0.4},{"t":6,"x":0.7,"y":0.4}] — so the area follows a subject that moves. Values above 1 are read as pixels instead of fractions. Returns an mp4 at the source size, frame rate and duration with the audio kept.

When to use — Hiding something in a video: a face, a number plate, a screen, a document. Areas are fractions of the frame, each with its own shape, style, optional time window and optional motion path, so a blur can follow a moving subject.

Trim a video or audio file: keep the part between a start and an end time and drop the rest. Use it to cut off an intro, cut off an ending, or pull one clip out of the middle. Pass the input as file_url or file_base64 + filename. 'start' is where the kept part begins and 'end' is where it stops — both accept seconds (12.5) or a clock time ('1:05', '00:01:05.250'). Omit start to keep from the very beginning; omit end to keep through to the end of the file; give at least one of them. 'duration' is an alternative to end (keep that many seconds from start) and is ignored if end is given. An end past the file length is clamped to the real end rather than failing. When the cut starts at 0 — or the input is audio-only — nothing is re-encoded: the result is bit-for-bit the source, produced almost instantly, and keeps the container that was uploaded. A cut starting between keyframes is re-encoded so it begins on exactly the requested frame, and a video re-encode comes back as MP4. A lossless cut ends on a packet boundary, so it can run up to ~0.15s longer than asked.

When to use — Cutting a video or audio file down: removing an intro or outro, or pulling one clip out of the middle. Give start and end, or start and duration.

Return the catalogue of input/output formats Convert Fleet supports, grouped by category, so you can choose a valid target_format for convert_file. Optionally pass 'from' to echo whether a given source format is recognised.

When to use — Before spending a credit on a conversion you are not sure about. Pass `from` to see exactly what one input format can become.

Return the caller's Convert Fleet plan (free/premium/enterprise) and remaining credits for the current period.

When to use — Checking the plan, the credits left in this period and the maximum size of a single file — do this BEFORE uploading something large, and when a call fails with -32001 so you can say whether it was the key or the quota.

Split a PDF into one file per page. Each page comes back as its own short-TTL signed download URL.

When to use — The user wants every page of a PDF as its own file.

Break a long HTML document into separate files — by heading level, by a custom delimiter, or by page break. Each section comes back as its own signed download URL.

When to use — A single large HTML export needs to become one file per chapter or section.

Render a live web page (by URL) or a string of HTML into a PDF, the way a browser would draw it.

When to use — Archiving a web page, turning a generated HTML report or invoice into a PDF, or capturing a page that needs CSS and JavaScript to look right.

Extract the picture from the END of a video as an image — the very last frame, or a chosen number of seconds before the end.

When to use — Making a closing thumbnail, checking how a render finished, or grabbing the final state of a screen recording.

Make a video last an exact number of seconds. New in-between frames are synthesised by motion interpolation, so a stretched clip still moves smoothly instead of stuttering on held frames; audio is tempo-stretched at its original pitch.

When to use — The user states a duration — 'make this 8-second clip run for 40 seconds', 'fit this into a 15-second slot' — or a speed multiplier.